Overlay two images on top of each other with rotation

I have two images that I would like to overlay on top of each other. The first is an outline of a human. The second is the outline of an arm. I would like to first rotate the image of the arm to a specific angle (say 135 degree), and then overlay it on top of the image of the human outline. That way it looks like the human is reaching for something on a shelf above him.

Answer 1

[Walter Roberson]

Read the images.

imcomplement the arm so that the dark outlines become bright and the bright outlines become dark. imrotate the complemented image. Now imcomplement that. You now have a rotated dark arm on a white background.

The reason to do the imcomplement before the rotation and then to complement back again, is that imrotate always adds a black background for parts of the new matrix that were originally outside the box. By making the background black, then rotating, then flipping colors, the effect is white background for the rotated arm.

Now, you have to figure out where exactly you want the rotated arm to go relative to the original. That would be a bit tricky to calculate unless you want to come up with some kind of mathematical formula such as "83% up from the bottom of the figure". It would probably be easier to just measure where you want it and use constants in the code.

Now figure out how high and wide an array you would need to join the body and the rotated arm together in one array. You will need to examine the rotation angle to figure out which side of the rotated arm is the attachment point to know whether the rotated arm goes to the left or the right of the body.

Once you have figured out the size needed for the composite image, pad the two images out to be the same size -- the padding will be different for the two images.

Now take min() of the padded images.

The result should be the composite image.

[Josh Tome]

Hi Walter,

Thank you for the response. After I posted the question I continued to do some digging on the forums and found the rotateAround function on the file exchange. I was able to get everything to work the way I want using that.

I do have a minor irritation with the result though. When I use the "imfuse" function to place the arm on top of the standing human image, the arm becomes a bit transparent so that you can see the outline of the body and head (depending on the arm position) behind it. Ideally I'd like the arm to block anything that is behind it.

Do you have suggestions on how to accomplish this?

A = imread('Full Standing Human without Arm.png');
B = imread('Human Arm Only Colored and Line.png');

% center of the shoulder joint (measured down and to the right from the top left corner of the image)
X = 41;
Y = 260;

% Made up angle to test code (degrees)
FlexionAngle = 140;

C=rotateAround(B,Y,X,FlexionAngle); % Rotates image around a point rotateAround(image,y-point,x    -point,degree)
D=~rotateAround(true(size(B)),Y,X,FlexionAngle); % Determines which pixes are black after the rotation of the image
C(D&~imclearborder(D)) = 255; % Turns black pixels white that were created after the rotation of the image

E=imfuse(A,C,'blend'); % plots both images on top of each other
F=imadjust(E,[],[],2.5); % adjusts the contrast of the image to make the outline darker
figure = imshow(F); % shows the resultant figure

[DGM]

1. Create a mask which defines the transparency of the foreground image.
2. Apply the same transformations to the mask as to the foreground image.
3. Compose the images.

In a way, you almost already have 1 and 2. Imfuse() is not a composition tool. It's not useful here.

If you add transparency to your arm image (see arm_rgba.png as an example), you could use your rotation tools on it as you see fit, so long as you also rotate the alpha array. When you're done, just compose the two images.

[FG,~,alpha] = imread('arm_rgba2.png'); % this image has transparency
BG = imread('Full Standing Human without Arm.png');

% do the composition
alpha = im2double(alpha);
outpict = im2uint8(alpha.*im2double(FG) + (1-alpha).*im2double(BG));

imshow(outpict,'border','tight')

[Josh Tome]

Does the transparency of my arm image have to be done in another software, or can this be accomplished in MATLAB?

Before starting this post I did attempt to make the background of the arm image transparent in a different software (Paint 3D), but when I joined the images I ended up with black as the background of the arm image instead of it being transparent.

[DGM]

I don't know what you did to create, read, or use the transparency, so I can't say why it was black. I can say that if you tried to read a transparent image via imread() with only a single output argument, then the transparent regions will be (typically) filled with black. Reading the alpha content requires using the multiple outputs as I showed. I don't know if that's what you're doing or not.

As far as doing it in MATLAB or externally, you could do it either way. I just did it externally.